Can it forget?
Interesting. This might be why a lot of tools aren’t currently being released in the EU? Pesky ‘Right to be forgotten/erasure’ eh?
Making AI models ‘forget’ undesirable data hurts their performance | TechCrunch
A new study finds that making generative AI models ‘forget’ information hurts their performance on other tasks, making them less useful.
Kyle Wiggers
“But current unlearning techniques are a double-edged sword: They could make a model like OpenAI’s GPT-4o or Meta’s Llama 3.1 405B much less capable of answering basic questions.”
Better to spin it as an issue with the EU, rather than the tech can’t handle it.
How did Meta put it?
“We will release a multimodal Llama model over the coming months, but not in the EU due to the unpredictable nature of the European regulatory environment,”
A lot of folk went ‘Wha?’. But this paper feels like a better explanation.